This is the mail archive of the
mailing list for the GCC project.
Re: [C++ PATCH]:Fix 14397
- From: Mark Mitchell <mark at codesourcery dot com>
- To: Nathan Sidwell <nathan at codesourcery dot com>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Fri, 05 Mar 2004 10:16:49 -0800
- Subject: Re: [C++ PATCH]:Fix 14397
- Organization: CodeSourcery, LLC
- References: <4048AA69.email@example.com>
Nathan Sidwell wrote:
this fixes 14397 a fallout from your 12226 patch. I'm mildly concerned
that we're building an IDENTITY_CONV for something that's really a
QUAL_CONV. I think this is another case of us not doing the algorithm
in the std, but doing one nearly the same that (presumably) has the same
built and tested on i686-pc-linux-gnu, ok?
I don't think that's probably quite the right fix. (For instance, if it
makes sense to add "const" it probably makes sense to add "volatile" too.)
The problem is probably that the result of build_dummy_object() is being
lvalue, but the original expression for which it is a proxy is an rvalue.
I think that we should wrap the dummy object in a NON_LVALUE_EXPR in
that case, so that the overloading routines will just DTRT.
Do you want to give that a try, or do you want to punt it back to me?